tech
February 6, 2026
AI companies want you to stop chatting with bots and start managing them
Claude Opus 4.6 and OpenAI Frontier pitch a future of supervising AI agents.
TL;DR
- Anthropic and OpenAI simultaneously released products focused on managing teams of AI agents, shifting the paradigm from conversational AI to a delegated workforce.
- Anthropic's Claude Opus 4.6 introduces 'agent teams' allowing multiple AI agents to split, coordinate, and execute tasks concurrently, suitable for tasks like codebase reviews.
- OpenAI launched Frontier, an enterprise platform where AI agents are given identities, permissions, and memory to connect with business systems, envisioned as 'AI co-workers'.
- Despite marketing as co-workers, these agents currently function best as tools requiring human supervision and course-correction.
- OpenAI also released a new macOS desktop app for Codex, acting as a 'command center for agents,' and GPT-5.3-Codex, an AI model that reportedly accelerated its own development.
- GPT-5.3-Codex outperformed Anthropic's Opus 4.6 on the Terminal-Bench 2.0 coding benchmark.
- The user's role is evolving from prompt-giver to supervisor, delegating tasks, monitoring progress, and intervening when necessary.
- Claude Opus 4.6 offers a large context window (up to 1 million tokens) and claims to outperform competitors on various benchmarks, though new OpenAI models remain competitive.
- These releases coincided with significant market volatility in software stocks, reportedly due to fears of AI models competing with established SaaS vendors by offering complete workflows.
- The concept of 'vibe working,' where AI handles tasks, is emerging as a parallel to 'vibe coding'.
Continue reading
the original article